\documentclass[12pt,a4paper]{article}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\usepackage{amssymb}
\usepackage{amsfonts}
\usepackage{amsmath}
\usepackage[doublespacing]{setspace}
\usepackage[margin=1in,nohead,letterpaper]{geometry}
\usepackage{geometry}
\usepackage{dcolumn}
\usepackage{color}
\usepackage{graphicx}
\usepackage{subcaption}
\usepackage{hyperref}
\usepackage{xcolor}
\setcounter{MaxMatrixCols}{10}
\renewcommand{\figurename}{Fig.}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{document}

\begin{center}
\Large\textbf{{Automation trends and labor markets}}
 \bigskip

\end{center}


\textbf{Irene Brambilla}, CEDLAS--IIE--FCE--UNLP and CONICET

\medskip

\textbf{Andrés César}, CEDLAS--IIE--FCE--UNLP and CONICET

\medskip

\textbf{Guillermo Falcone}, CEDLAS--IIE--FCE--UNLP and CONICET

\medskip

\textbf{Leonardo Gasparini}, CEDLAS--IIE--FCE--UNLP and CONICET

 \bigskip
 
 
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\newpage
\setcounter{page}{1}

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{table}[!t]
\begin{center}
\caption{UNIDO industry data}
\label{tab_unidodesc}
\input{./results/table1.tex}
\begin{minipage}{14cm}
\scriptsize\rule{0cm}{0.55cm}Notes: Own calculations from UNIDO. Columns (1) and (2) show the number of observations. Columns (3) to (6) show the average labor share, defined as industry wagebill over industry value added, for different time periods. The last line computes the average across all countries in the table except Argentina.
\end{minipage}
\end{center}
\end{table}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{table}[!t]
\begin{center}
\caption{SEDLAC individual and district level data}
\label{tab_sedlacdesc}
\input{./results/table2.tex}
\begin{minipage}{16cm}
\scriptsize\rule{0cm}{0.55cm}Notes: Own calculations from SEDLAC database. Columns (1) to (3) show the number of observations at the individual level, at the district level, and the number of years of data per country. Columns (4) to (7) report average statistics during 1992--2015. Labor market statistics are restricted to adults aged 18--65. Unemployment is the share of adults in the labor force that have been actively looking for a job in the last month. Labor income is the monthly value expressed in constant USD PPP 2011. Poverty rate is the percentage of population with income below the official moderate poverty line. The ratio p75/p25 is the inter-quartile range of labor income.
\end{minipage}
\end{center}
\end{table}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{table}[!t]
\begin{center}
\caption{Industries and occupations with high and low levels of routinization}
\label{tab-rtc}
\input{./results/table3.tex}
\par
\begin{minipage}{15cm}
\scriptsize\rule{0cm}{0.55cm}Notes: Own calculations from PIAAC surveys. The first panel shows the three industries with highest and lowest levels of routinization, computed as in equation (\ref{eq-rtc2}). The second panel shows the three occupations with highest and lowest levels of routinization, computed as in the first average in equation (\ref{eq-rtc3}). $RTC_6$ is computed from DOT data.
\end{minipage}
\end{center}
\end{table}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{figure}[!t]
\caption{Dispersion of RTC index across districts}
\label{fig_hist2}

~

\centering
\includegraphics[width=3.5in]{./results/figure1.png}

\par
\begin{minipage}{13cm}
\scriptsize \rule{0cm}{0.35cm}Notes: Histogram shows the frequency distribution of the RTC index across districts based on the occupation composition of the initial year of data. Source: SEDLAC and PIAAC.
\end{minipage}
\end{figure}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{figure}[!t]
\caption{Trends in adoption of automation technology}
\label{fig_robots1}
~
\par
\centering
\par
\includegraphics[width=2in]{./results/figure2-Argentina.png}
\includegraphics[width=2in]{./results/figure2-Brazil.png}
\includegraphics[width=2in]{./results/figure2-Chile.png}

\includegraphics[width=2in]{./results/figure2-Colombia.png}
\includegraphics[width=2in]{./results/figure2-Mexico.png}
\includegraphics[width=2in]{./results/figure2-Peru.png}
\par
\begin{minipage}{14.5cm}
\scriptsize \rule{0cm}{0.35cm}Notes: total number of industrial robots. Source: International Federation of Robotics (IFR).
\end{minipage}
\end{figure}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{table}[!t]
\begin{center}
\caption{Industry level regressions}
\label{tab_reg1a}
\small{\input{./results/table4.tex}}
\par
\begin{minipage}{15cm}
\scriptsize\rule{0cm}{0.55cm}Notes: Dependent variables are: columns (1) and (2) log industry employment, columns (3) and (4) log industry average wage, columns (5) and (6) industry share of labor. Table shows coefficients $\alpha_1$ and $\alpha_2$ from regression equations (\ref{reg1}) and (\ref{reg2}). Columns (1), (3), (5) refer to changes in outcome defined as $t_2 - t_1$. Columns (2), (4), (6) refer to changes in outcome defined as $t_3 - t_2$. Regressions control for initial labor share, initial log value added per worker, and change in log value added per worker. Results are robust to different combinations of the control variables. Robust standard errors in parenthesis. Significance at the 1, 5 and 10 percent levels denoted with ***, ** and *. Source: UNIDO and PIAAC.
\end{minipage}
\end{center}
\end{table}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{table}[!t]
\caption{Industry level regressions. \\Parameterization of technology adoption}
\label{tab_reg3a}
\begin{center}
\small{\input{./results/table5.tex}}
\par
\begin{minipage}{15cm}
\scriptsize\rule{0cm}{0.55cm}Notes: Dependent variables are: log industry employment, log industry average wage, industry labor share of value added. Table shows coefficients $\alpha_4$ from regression equation (\ref{reg4}). Regressions control for initial labor share, initial log value added per worker, and change in log value added per worker. Robust standard errors in parenthesis. Significance at the 1, 5 and 10 percent levels denoted with ***, ** and *. Source: UNIDO, PIAAC and IFR.
\end{minipage}
\end{center}
\end{table}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{table}[!t]
\begin{center}
\caption{Industry level regressions.\\Share of industry in total employment}
\label{tab_reg1b}
\small{\input{./results/table6.tex}}
\par
\begin{minipage}{15cm}
\scriptsize\rule{0cm}{0.55cm}Notes: In columns (1) and (2) the dependent variable is industry share in total employment for the years 2005 and 2013. Independent variable: routine task content index. Regressions control for country effects. Columns (3) and (4) are analogous to Table \ref{tab_reg1a} with dependent variable industry share in total employment. Regressions control for initial labor share, initial log value added per worker, and change in log value added per worker. Robust standard errors in parenthesis. Significance at the 1, 5 and 10 percent levels denoted with ***, ** and *. Source: UNIDO and PIAAC.
\end{minipage}
\end{center}
\end{table}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{table}[!t]
\begin{center}
\caption{District level regressions. Employment and wages}
\label{tab_llm1}
\small{\input{./results/table7.tex}}
\par
\begin{minipage}{15cm}
\scriptsize\rule{0cm}{0.55cm}Notes: Dependent variables are: columns (1) and (2) district employment rate, columns (3) and (4) district unemployment rate, columns (5) and (6) log district average wage. Table shows coefficients $\phi_1$ and $\phi_2$ from regression equations (\ref{reg5}) and (\ref{reg6}). Columns (1), (3), (5) refer to changes in outcome defined as $t_2 - t_1$. Columns (2), (4), (6) refer to changes in outcome defined as $t_3 - t_2$. Regressions control for initial average wage and employment rate. Robust standard errors in parenthesis. Significance at the 1, 5 and 10 percent levels denoted with ***, ** and *. Source: SEDLAC and PIAAC.
\end{minipage}
\end{center}
\end{table}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%   
\begin{table}[!t]
\begin{center}
\caption{District level regressions.\\
Parameterization of technology adoption}
\label{tab_llm3}
\small{\input{./results/table8.tex}}
\par
\begin{minipage}{15cm}
\scriptsize\rule{0cm}{0.55cm}Notes: Dependent variables are: Employment rate, unemployment rate, log district average wage. Regressions control for initial average wage and employment rate. Robust standard errors in parenthesis. Significance at the 1, 5 and 10 percent levels denoted with ***, ** and *. Source: SEDLAC, PIAAC and IFR.
\end{minipage}
\end{center}
\end{table}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{table}[!t]
\begin{center}
\caption{District level regressions.\\Unemployment by skill groups}
\label{regllm3-c}
\small{\input{./results/table9.tex}}
\par
\begin{minipage}{15cm}
\scriptsize\rule{0cm}{0.55cm}Notes: Dependent variable: unemployment rate computed for three skill groups. Unskilled workers: no high school degree. Semi-skilled workers: high school degree and no further education. Highly skilled workers: tertiary education or university degree. Table shows coefficients $\phi_1$ and $\phi_2$ from regression equations (\ref{reg5}) and (\ref{reg6}). Columns (1), (3), (5) refer to changes in outcome defined as $t_2 - t_1$. Columns (2), (4), (6) refer to changes in outcome defined as $t_3 - t_2$. Regressions control for initial average wage and employment rate. Robust standard errors in parenthesis. Significance at the 1, 5 and 10 percent levels denoted with ***, ** and *. Source: SEDLAC and PIAAC.
\end{minipage}
\end{center}
\end{table}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{table}[!t]
\begin{center}
\caption{District level regressions.\\Labor informality by skill groups}
\label{regllm3-d}
\small{\input{./results/table10.tex}}
\par
\begin{minipage}{15cm}
\scriptsize\rule{0cm}{0.55cm}Notes: Dependent variable: labor informality rate is the share of salaried workers with no contributions to the pension system. It is computed for three skill groups. Unskilled workers: no high school degree. Semi-skilled workers: high school degree and no further education. Highly skilled workers: tertiary education or university degree. Table shows coefficients $\phi_1$ and $\phi_2$ from regression equations (\ref{reg5}) and (\ref{reg6}). Columns (1), (3), (5) refer to changes in outcome defined as $t_2 - t_1$. Columns (2), (4), (6) refer to changes in outcome defined as $t_3 - t_2$. Regressions control for initial average wage and employment rate. Robust standard errors in parenthesis. Significance at the 1, 5 and 10 percent levels denoted with ***, ** and *. Source: SEDLAC and PIAAC.
\end{minipage}
\end{center}
\end{table}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{table}[!t]
\begin{center}
\caption{District level regressions. Poverty and inequality}
\label{regllm3-a}
\small{\input{./results/table11.tex}}
\par
\begin{minipage}{15cm}
\scriptsize\rule{0cm}{0.55cm}Notes: Analogous to Table \ref{tab_llm1} with outcome variables: columns (1) and (2): Poverty rate computed using per capita family income; columns (3) and (4): inter-quartile range (ratio p75/p25) of labor income. Regressions control for initial poverty rate and inter-quartile ratio of income. Robust standard errors in parenthesis. Significance at the 1, 5 and 10 percent levels denoted with ***, ** and *. Source: SEDLAC and PIAAC.
\end{minipage}
\end{center}
\end{table}

%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\newpage
\clearpage
\section*{Appendix: Data and Additional Results}
\setcounter{figure}{0} \renewcommand{\thefigure}{A\arabic{figure}}
\setcounter{table}{0} \renewcommand{\thetable}{A\arabic{table}}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\clearpage

\begin{figure}[!ht]
\caption{Correlation of definitions of RTC indexes}
\label{fig_piaaccorr1}
~
\par
\centering
\subcaption{Industry indexes}
\includegraphics[width=2in]{./results/figureA1-h2.png}
\includegraphics[width=2in]{./results/figureA1-h3.png}
\includegraphics[width=2in]{./results/figureA1-h4.png}
\includegraphics[width=2in]{./results/figureA1-h5.png}
\includegraphics[width=2in]{./results/figureA1-h6.png}

~

\subcaption{Occupation indexes}
\includegraphics[width=2in]{./results/figureA1-g2.png}
\includegraphics[width=2in]{./results/figureA1-g3.png}
\includegraphics[width=2in]{./results/figureA1-g4.png}
\includegraphics[width=2in]{./results/figureA1-g5.png}
\includegraphics[width=2in]{./results/figureA1-g6.png}

\par
\begin{minipage}{16cm}
\scriptsize \rule{0cm}{0.35cm}Notes: Figure plots the correlation between industry-level indexes $RTC_1$ and $RTC_2$ to $RTC_6$ (top panel), and the correlation between occupation-level indexes $RTC_1$ and $RTC_2$ to $RTC_6$ (bottom panel). Indexes $RTC_1$ to $RTC_5$ are computed from the pooled PIAAC surveys, whereas $RTC_6$ is computed from the index of Autor and Dorn (2013).
\end{minipage}
\end{figure}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{figure}[!h]
\caption{Dispersion of RTC index across districts}
\label{fig_hist3}

~

\centering
\includegraphics[width=2in]{./results/figureA2-3.png}
\includegraphics[width=2in]{./results/figureA2-4.png}
\includegraphics[width=2in]{./results/figureA2-5.png}
\includegraphics[width=2in]{./results/figureA2-6.png}
\includegraphics[width=2in]{./results/figureA2-7.png}
\includegraphics[width=2in]{./results/figureA2-8.png}
\includegraphics[width=2in]{./results/figureA2-9.png}
\par
\begin{minipage}{14.5cm}
\scriptsize \rule{0cm}{0.35cm}Notes: Histograms show the frequency distribution of the $RTC_1$ index across districts.
\end{minipage}
\end{figure}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\clearpage
\begin{table}[!t]
\begin{center}
\caption{PIAAC surveys}
\label{tab_piaacdesc}
\input{./results/tableA1.tex}
\begin{minipage}{13cm}
\scriptsize\rule{0cm}{0.55cm}Notes: Table shows the percentage of individuals that respond ``yes'' to performing six flexible tasks often (Supervising, Planning, Solving problems, Producing written output, Giving presentations or sales pitches, Calculating budgets), the average of the four flexibility indexes across individuals ($F_1$, $F_2$, $F_3$, $F_4$, $F_5$), and the number of observations. Calculations are based on employed individuals that can be matched to an ISCO 08 occupation.
\end{minipage}
\end{center}
\end{table}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{table}[!ht]
\begin{center}
\caption{Correlation of RTC indexes computed from different samples}
\label{tab_piaaccorr1}
\input{./results/tableA2.tex}
\par
\begin{minipage}{13cm}
\scriptsize\rule{0cm}{0.55cm}Notes: RTC indexes are computed from different samples at the occupation level (Panel A) and at the industry level (Panel B). Column 1 displays the correlation of $RTC_1$ computed from surveys of the four Latin American countries pooled together, with the RTC index computed separately from the survey of each of the four Latin American countries. Columns 2, 3, 4 and 5 compute analogous correlations for $RTC_2$, $RTC_3$, $RTC_4$ and $RTC_5$. To compute the correlations we keep occupations and industries with at least 25 observations for each of the four Latin American countries.
\end{minipage}
\end{center}
\end{table}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{table}[!ht]
\begin{center}
\caption{Industry level RTC indexes}
\label{tab_indrtc1}
\input{./results/tableA3.tex}
\par
\begin{minipage}{15cm}
\scriptsize\rule{0cm}{0.55cm}Notes: RTC indexes $RTC_1$ to $RTC_5$ are computed from pooled PIAAC surveys as weighted averages of the individual level flexibility indexes $F_1$ to $F_5$. $RTC_6$ is computed as a weighted average of the occupation level index of Autor and Dorn (2013), using the occupation shares in industry employment as weights. Column (7) displays the number of surveyed individuals that report employment in each industry in the PIAAC surveys. Columns (1) to (5) are computed based on the number of observations in the PIAAC surveys in column (7).
\end{minipage}
\end{center}
\end{table}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{table}[!ht]
\begin{center}
\caption{Occupation level RTC indexes}
\label{tab_occrtc1}
\small{\input{./results/tableA4.tex}}
\par
\begin{minipage}{16cm}
\scriptsize\rule{0cm}{0.55cm}Notes: RTC indexes $RTC_1$ to $RTC_5$ are computed from pooled PIAAC surveys as weighted averages of the individual level flexibility indexes $F_1$ to $F_5$. $RTC_6$ is the occupation level index of Autor and Dorn (2013). Column (7) displays the number of surveyed individuals that report employment in each occupation in the PIAAC surveys. Columns (1) to (5) are computed based on the number of observations in the PIAAC surveys in column (7).
\end{minipage}
\end{center}
\end{table}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{table}[!ht]
\begin{center}
\caption{Average RTC by country}
\label{tab_ctry_rtc}
\input{./results/tableA5.tex}
\par
\begin{minipage}{11cm}
\scriptsize\rule{0cm}{0.55cm}Notes: Table shows the weighted average of district-level RTC indexes for each country.
\end{minipage}
\end{center}
\end{table}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{table}[!t]
\begin{center}
\caption{Industry level regressions}
\label{tab_off1}
\small{\input{./results/tableA6.tex}}
\par
\begin{minipage}{11cm}
\scriptsize\rule{0cm}{0.55cm}Notes: Analogous to Table \ref{tab_reg1a} adding a control for offshoring.
\end{minipage}
\end{center}
\end{table}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\begin{table}[!t]
\begin{center}
\caption{Industry level regressions}
\label{tab_off2}
\small{\input{./results/tableA7.tex}}
\par
\begin{minipage}{11cm}
\scriptsize\rule{0cm}{0.55cm}Notes: Analogous to Table \ref{tab_reg3a} adding a control for offshoring.
\end{minipage}
\end{center}
\end{table}
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
\end{document}
